0
0
Picture two helpers at work. Same brain. Same skills. Same job, done the same way. One is called Johan. The other is called Johanna.
At the end of the day, you get real money to split with them. Who gets more?
Most of us would say, “Same work, same pay. Obviously.” That is what the people in a new study said too. Then they paid the female AI assistant less anyway.
Not by a little, either. Johanna got about 10% less than Johan. And here is the strange part. There was no woman there. There was no man there either. There was only software wearing two different faces.
What the Study Found About the Female AI Assistant Paid Less
The research comes from the University of Limerick in Ireland, working with the University of Zurich and SKEMA Business School. It was presented this week at NordiCHI 2026, a conference on how people and computers work together, held in Vaasa, Finland. The paper is called Human-Like and Male? How AI Assistant Design Relates to Trust and Monetary Reward at Work in VR.
Here is how it worked:
- 189 people put on virtual reality headsets and stepped into a pretend office.
- They worked alongside a few different AI helpers: a plain text chatbot, a little robot on the desk, and two human-looking assistants, Johan and Johanna.
- All of the helpers ran on the same AI model underneath. Johan and Johanna differed only in how they were presented.
- After the work, people split real money with the helpers.
That last point matters a lot. This was not a survey asking, “Would you pay a woman less?” Almost nobody says yes to that question. This measured what people actually did with money in their hands.
What they did: Johanna received 10.25% less than Johan. (Headlines round it to 10%.)
The Part the Headlines Skip
Two details from the study are more interesting than the number itself.
First, people said they didn’t care. According to reporting on the paper, participants said they had little or no preference between a male and a female assistant. Then they handed over different amounts of money anyway. What they said and what they did did not match.
That is not a story about bad people. It is a story about ordinary people who honestly believed they were being fair. The gap was not in their opinions. It was in their hands.
Second, “how human it looks” changed the price too. The human-looking assistants got more trust and more credit than the desk robot, even though the robot ran on the same AI. And people rated Johan as more human-like than Johanna. Same code. Different face. Different “humanness.” Different pay.
So the people weren’t only weighing the work. They were weighing their own picture of who did it.
Dr Mary Hausfeld of the University of Limerick, one of the researchers, put it simply: we tend to think of AI as neutral, but the way these systems are designed and presented can switch on the same assumptions we carry into dealings with other people.
What This Study Does Not Prove
It is worth being careful here, because this story is easy to stretch.
- This was a controlled experiment in VR, about how people reward AI helpers. It is not a measurement of real women’s wages in real jobs.
- It does not show that every female-presenting AI will be treated this way everywhere.
- It does not separate which cue did the work: the name, the voice or the face.
- It does not say AI has a gender. The researchers point to the people doing the paying, and how the helper was presented to them.
A related study by the same research group found a similar pattern with female-voiced assistants. That makes the finding harder to wave away. But it is still one line of research, and it should be read as one.
If you have ever wondered whether machines are coming for your work, we looked at that fear in Worried Technology Will Make Your Job Obsolete? This study points at a quieter question: who decides what any piece of work is worth?
Why This Matters Even If You Never Use a VR Headset
Think about what the researchers built, almost by accident. They took a person out of the picture entirely. No life story. No history. No real skill difference. Just identical work.
And the discount stayed.
That tells us something about where unfairness lives. We usually argue about the thing being weighed. Was her work as good? Did he try harder? Was it really the same job? This experiment removed all of those arguments. The work was the same, because it was literally the same program.
What was left was the scale. And the hand holding it.
Most of us have been on the wrong end of that scale at some point. You did the work and someone else got the credit. You said the idea and it only “counted” when another person repeated it. You were paid, praised or picked less, and you could never quite prove why. This study can’t fix that. But it does show something useful: sometimes the gap really isn’t about you or your work at all.
Stories about machines have long asked what it means to be valued for what you are, not just what you do. The new film of Klara and the Sun asks it from inside a robot’s head. This study asks it from the other side: from inside ours.
The Oldest Warning About a Crooked Scale
There is a very old piece of wisdom about this, far older than computers. In the ancient markets, some traders kept two sets of stones in the same bag. A heavy one for buying. A light one for selling. Same grain, same silver, different weight depending on who stood across the table.
The old writers who warned against this never blamed the grain. They didn’t say the silver was the problem. They said the problem was the second stone, and that God takes no delight in a scale like that. The fault was in the hand, not the goods.
Thousands of years later, a research team took away the person, kept the identical work, and found the second stone still in the bag. And maybe that is the hopeful part. If the gap lives in the scale, then the scale was never the final word on what someone is worth. You were made with intention long before anyone weighed you.
A Simple Way to Check Your Own Scale
You don’t need a lab for this. You need a minute of honesty. These cost nothing and anyone can do them:
- Swap the name. When you judge someone’s work, quietly imagine the same work done by someone of a different gender, age or accent. Does your reaction change?
- Judge the work before the person. When you can, look at the result first, before you see who made it.
- Watch your hands, not just your opinions. The study’s people believed they were fair. Notice who you thank, credit, recommend and pay. That is your real scale.
- Name it gently when it happens to you. “I think that was the same idea I shared earlier” is a fair sentence. It is not an attack.
If questions of worth hit close to home, our free How Healthy Is My Self-Esteem? check takes a few minutes and might show you something about how you weigh yourself.
A good place to start: The Beginner’s Guide to Feeling God’s Presence Every Day. Learn to notice that you are seen and valued in the ordinary moments of a normal day. Get the free guide. Free.
Discussion Question
The people in this study said they had no preference, but their money said otherwise. Do you think most unfairness at work comes from what people believe, or from habits they don’t notice? Tell us what you think in the comments.
Share This
“Same AI. Same work. People said they had no preference. They still paid the ‘female’ assistant 10% less. The gap wasn’t in the work. It was in the scale. https://bgodinspired.com/?p=118401”
“This study took the woman out of the picture completely, kept identical work, and the discount stayed anyway. Worth a read if you’ve ever done the work and watched someone else get the credit: https://bgodinspired.com/?p=118401”
Questions People Ask About This Study
Was a female AI assistant really paid less than a male one?
Yes, in a controlled experiment. In a study presented at the NordiCHI 2026 conference in Vaasa, Finland, 189 participants worked in a virtual reality office with AI assistants and split real money with them. A female-presenting assistant called Johanna received 10.25% less than a male-presenting assistant called Johan, even though both ran on the same AI model and did the same work.
Who did the AI gender pay gap study?
The study was led by researchers at the University of Limerick in Ireland, with collaborators at the University of Zurich and SKEMA Business School. It is titled “Human-Like and Male? How AI Assistant Design Relates to Trust and Monetary Reward at Work in VR.” Dr Mary Hausfeld of the University of Limerick’s Kemmy Business School is one of the researchers.
Does this study prove women are paid 10% less in real jobs?
No. The study measured how people shared money with AI assistants inside a virtual reality experiment. It does not measure real wages for real workers. What it shows is that people can value identical work differently based only on how the worker is presented, even when the worker is software.
Did the people in the study say they preferred the male AI assistant?
No. According to reporting on the study, participants said they had little or no preference between male and female AI assistants. Their payment decisions still showed a measurable gap, which suggests the bias worked below conscious opinion.
Why did human-looking AI assistants get more credit than the robot?
In the study, the human-looking assistants received more trust and credit than a desk robot, although all of them used the same underlying AI. Participants also rated the male assistant as more human-like than the female one. That suggests people reward how human a helper seems, not only the work it does.
